I have three things that come to my mind:

1. What are you comparing to your GPS program that makes you think it 
is not accurate? If it is your car's odometer, they can be off by 
several percent I think.

2. If you are using your car's odometer, are you considering that the 
car will have to drive farther over the hills (the vertical component) 
than the GPS will register?

3. You aren't comparing "as the crow flies" distance to your driven 
distance by odometer, are you?

On Monday, July 26, 2004, at 07:14 PM, Edwards, Waverly wrote:

> I'm working on a GPS program and I'm having difficulty calculating 
> distance.
> Using the distance between the longitude/latitude moved is not an 
> accurate
> means for me to discover how far I have moved.  In some instances it 
> is correct
> but accumulative distance over a distance of ten miles moved is at 
> times off by
> 3/4 of a mile.  In a twenty mile drive my distance may be off by two 
> tenths of a
> mile.  In short distances lets say over the course of a two miles my 
> readings
> vary wildly.
>
> My GPS samples are taken once a second so I could possibly use a 
> distance
> over time calculation.  My distance in miles per hour [seem] to be 
> more consistent
> but I don't know if I can trust that.  GPS's aren't perfect.  My older 
> Garmin III+ has
> a limit of 5 meter accuracy (roughly 15 feet) but errors as much as 15 
> meters
> (roughly 45 feet) where the norm is about 18-21 feet.
>
> I've been trying to figure out how to factor in approximate errors in 
> position that are
> given as HDOP (Horizontal Dilution Of Position) without success.
>
> What I'd like to do is some combination of guestimating how far I've 
> moved as an
> average of calculating the distance in distance moved using 
> longitude/latitude and
> speed over time.  On an older model GPS I'd like to see if I can get 
> within 100 feet
> of my mark while driving.
>
> I could lower my sampling rate of position and calculate distance over 
> a five second
> versus a one second rate which would lower my error by twenty percent 
> and try to
> project (interpolate) how far I am from my target.  I thought about a 
> moving
> average but I don't know how to do that.
>
> Any suggestions on how I might accomplish my mission?